The Agent Skills Directory

[COMMAND_EXECUTION]: The skill leverages the ae-cli tool for data retrieval. It follows a highly secure execution pattern by mandating that the agent must first discover project-specific metadata (event and property names) via ae-cli analysis-meta commands before attempting to execute data queries. This 'contract verification' prevents the agent from guessing field names or being misled by unexpected data structures.
[DATA_EXPOSURE]: While the skill accesses sensitive user data, it is the primary intended purpose of the tool. It includes a comprehensive 'Data Privacy & Security' section that defines mandatory masking for sensitive fields such as phone numbers, IP addresses, and real names. It also references audit logging and access control as required practices.
[INDIRECT_PROMPT_INJECTION]: The skill handles data from external CLI outputs, creating a theoretical attack surface. However, this is mitigated through a robust evidence chain:
Ingestion points: Data enters the context via ae-cli analysis entity-detail run and event-detail run outputs.
Boundary markers: The skill uses clear separators and titles (e.g., ━━━━━━━━━━━━━━━) to delimit data blocks in its output.
Capability inventory: The skill is limited to data retrieval, aggregation, and formatting; it does not perform recursive execution of data-derived commands.
Sanitization: The skill instructions explicitly require masking sensitive identifiers before presenting them to the user, reducing the risk of exfiltration via data poisoning.

single-user-inspector